There was an old television show called “Truth or Consequences.” I remember the name better than I remember much of the show’s content, but lately the phrase has taken on a different and even a critical meaning for me. Artificial intelligence (AI) is becoming remarkably good at helping us arrive at answers and decisions. Sometimes it does much more than help. It decides and substantially shapes where we end up. Yet when that techno-decision leaves the screen and enters our real world, something curious happens. The cognition may be shared, but the consequences aren’t.
Cognitive Effort and Accountability
Imagine making an important decision with the help of AI. You provide the context, ask some questions, and quickly come to a conclusion. But my bet, borrowed from the ubiquitous 80/20 rule, is that you contributed 20 percent of the actual cognitive work. Not 80 percent, but 20 percent. If the decision turns out brilliantly, you’re pleased. If it goes badly, however, the machine doesn’t get called into the boss’s office.
That asymmetry is fascinating to me and may be one of today’s most pressing issues. A human may contribute 20 percent of the cognition and still own 100 percent of the consequence.
Let’s start here. Cognition and consequence have typically been close neighbors that share a cognitive border. We rely on all sorts of tools, but the center of intellectual gravity usually remains somewhere inside the person making the decision. Think about a physician confronting a difficult diagnosis. Plenty of other people and sources might inform the decision, but eventually the physician stands accountable. AI complicates this because it doesn’t simply hand us information anymore. Increasingly, it participates in the reasoning that turns information into a decision.
To varying degrees, most of us have experienced this. We can begin with an idea or even an intuition, bring it into a conversation with a large language model (LLM). We iterate with AI and often find ourselves in a new and unexpected place. Who thought the final thought? I’m increasingly unsure that the question has a clean answer. Sometimes the cognition seems to emerge from the interaction itself.
Human Responsibility in Decisions Made With AI
But responsibility doesn’t become equally fuzzy. In a general sense, AI can’t lose its reputation or its job. It can’t legitimately apologize to someone harmed by a recommendation or even lie awake wondering how it got something so wrong.
AI can have enormous influence over the cognitive process and almost none of the consequence that follows.
And I think this gets harder as AI gets better. If a machine contributes 5 percent to my reasoning, calling the decision mine seems easy enough. What about 50 percent? What about 80? At some point, we may find ourselves holding people fully responsible for decisions they only partially thought through themselves. Perhaps that’s appropriate. Someone still chose to act. But I’m not sure the philosophy of responsibility is quite as simple as the organizational chart might suggest.
We’ve spent enormous energy asking whether machines can think, reason, understand, or possess something resembling intelligence. Those are fascinating questions. But another question begins after all that thinking is finished.
Who owns a decision when no single mind fully made it?
For now, we can give AI more and more of the cognitive work without giving it any meaningful share of what happens next. The thinking can become distributed across human and machine, while responsibility remains concentrated in one place: us.